In Live Optics, Peak Composite Graphs are different than Aggregation Graphs.
Aggregation graphs take similar values, SUMs or aggregates them and create a new value that represents the whole.
For example, Port 1 has 32 MB/s of throughput and Port 2 has 20 MB/s of throughput. 32+20 can be aggregated to show that the total throughput is 53 MB/s.
A Composite Graph will also be made up of parts that are related but need to be tracked individually. In other words, SUMming the values does not lead to a truth.
For example, if Port 1 is at 95% utilization and Port 2 is at 5% utilization, the array is not at 100% port utilization. Port 1 is in trouble and Port 2 is underutilized.
For this reason, we have created what we call a Peak Composite Graph and the rest of this write up will help you understand how it’s constructed.
Live Optics will plot a timeline of the sample data retrieved. Let’s assume that we have these 4 ports that are all doing a similar duty, but what matters is if any of the ports are exceeding our threshold of 80% utilization. If any port violates this this threshold, it’s a problem. From a dashboard perspective, it doesn’t even matter which one.
The Peak Composite Graph will look across all the objects being measured and report the Peak value to the Graph.
For illustrative proposes, the graph below shows the Peak Composite, meaning the highest Peak from all 4 ports is elevated to be the reporting value in the Composite graph. More than one port exceeding the threshold (or a tie) is not relevant. We know that at that time there was a utilization threshold exceeded on the array.
Hovering over the indicator in the Peak Composite Graph will tell you the threshold that was exceeded.
Fig 1: In the mocked up chart below you can see how the highest peak red bars are elevated to be represented in the overall Composite Graph.
The Peak Composite Graph will always show a fixed number of plot points. So no matter what the recording duration or sample rate, they will be fixed. Live Optics will use a roll up technique that is dynamic based on sample rate and sample duration to optimize the graph, but leave the outcome the same.
Example: If 10 samples are rolled into 1, then the highest sample value of the 10 samples would represent the set in the roll up.
Here is a real life example of a rather average sample set.
Here is another real life example where the samples indicate that attention might be needed.